![](/static/253f0d9b/assets/icons/icon-96x96.png)
![](https://lemmy.world/pictrs/image/41e39366-cb91-4d4a-bc07-a47621cb7d5f.jpeg)
It’s pretty obvious that the censorship training makes the AI less capable in general for a lot of use cases, even ones that don’t directly touch on nsfw content. It’s a major reason behind the popularity of local models. OpenAI must be feeling the pressure to compete even more than they’re afraid of people writing articles like this one at this point.
Explicit types are just laziness, you should be catching exceptions anyways.